From OntologPSMW

Jump to: navigation, search
[ ]
    (1)
Session Binil Starly
Duration 1 hour
Date/Time 08 April 2020 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CEST
Convener KenBaclawski
Track Use Cases

Contents

Knowledge graphs, closely related to ontologies and semantic networks, have emerged in the last few years to be an important semantic technology and research area. As structured representations of semantic knowledge that are stored in a graph, KGs are lightweight versions of semantic networks that scale to massive datasets such as the entire World Wide Web. Industry has devoted a great deal of effort to the development of knowledge graphs, and they are now critical to the functions of intelligent virtual assistants such as Siri and Alexa. Some of the research communities where KGs are relevant are Ontologies, Big Data, Linked Data, Open Knowledge Network, Artificial Intelligence, Deep Learning, and many others.     (2A)

Agenda     (2B)

  • Professor Binil Starly Building an Open Knowledge Network (OKN) Graph in Product Design & Manufacturing Slides Video Recording YouTube Video     (2B1)
    • Abstract: Try searching for “Precision gear manufacturers in North Carolina that have served aerospace industry” OR search for “Precision Gear 3D models” in any of the online 3D model galleries. None of these search terms retrieve any meaningful results, primarily because we lack tools to index and search through text and 3D model information. Access to 3D models and manufacturing services is severely limited if potential clients cannot find them. Can we leverage the power of machine learning and the broader AI algorithms to create an Open Knowledge Network integrating diverse data sources from millions of publicly available 3D CAD product models, manufacturing textbooks, blogs, videos, web articles, technical articles etc. to answer those two relatively simple queries? This talk will describe activities to the challenge of building a knowledge network graph in design and manufacturing with demo application that showcase how a user might benefit from it. Current solutions to classify and categorize 3D product model data are heavily dependent on human annotation and manual classification. Searching through 3D models must be multi-modal – how do we jointly search with text, image and 3D model data. Can we learn from global search engines which understand context around the content of a website and not rely simply on text scraping and indexing content? Understand the challenges and involve the technical community to build a Knowledge Graph for design and manufacturing data.     (2B1A)
    • Bio: Binil Starly heads the Digital Manufacturing group at NC State Industrial and Systems Engineering department. His laboratory is working on technologies that merge the digital and the physical world towards advancing both discrete and continuous manufacturing processes. His work is supported by the US National Science Foundation, Department of Energy and the Department of Defense. He has received the National Science Foundation CAREER award for research in engineering living tissue systems, the SME Young Manufacturing Engineer and other teaching awards. He started his career at the University of Oklahoma, grew from Assistant Professor to Associate Professor with tenure, spending 7 years with the university. He then worked for NC State University where he has been for the past 6 years and during this time, he was promoted to Full Professor. He currently holds the James. T. Ryan Professorship in recognition for his work on smart manufacturing technologies.     (2B1B)
  • Other links:     (2B2)
  • Discussion     (2B3)

Conference Call Information     (2C)

Attendees     (2D)

Discussion     (2E)

[12:05] David Eddy: great to open (pre meeting chatter) with "interoperable" rather than the delusional "integrated"     (2E1)

[12:13] Mike Bennett: This idea is readily extensible in to something I've been thinking about in terms of product 'hacks'. Is there a tool that does these things now, ontology driven?     (2E2)

[12:14] Ravi Sharma: Binil - trust is faster changing in emergencies due to QOS issues sometimes     (2E3)

[12:22] Ravi Sharma: Binil - when you mix and match #D components how do you ensure physical and dynamic aspects to ensure assembly works?(solvers used?)     (2E4)

[12:24] Mike Bennett: What viz is being used. Seems to be able to show a graphic for the individual thing rather than e.g. a colored box as you would see in Gruff and others. This tool would also be useful for graphical representation of information e.g. social media snippets. Is this a 3rd party library?     (2E5)

[12:24] David Eddy: What's the "bridge" between the simple, constrained world of 3D to the N-dimensional world of data?     (2E6)

[12:25] David Eddy: duly note on "standards"... I did not do the research but a reliable source told me in early industrial age UK, it took 75 years to agree on screw thread standards     (2E7)

[12:26] ToddSchneider: For the manual annotation process were any prompts or a priori 'labels' provided to aid in better annotations?     (2E8)

[12:26] David Eddy: and when I was last dealing in 3D things... there were 5 standard threads in the UK     (2E9)

[12:27] Mike Bennett: So we need an ontology of capability ,capacity, function etc. (similar to work at IOF on these concepts)     (2E10)

[12:28] David Eddy: IOF = ?     (2E11)

[12:28] Ravi Sharma: how do you dynamically update manufacturing and service providers data changes?     (2E12)

[12:29] David Eddy: @Ravi... folks are going to have to embrace pervasive & automated change control (SCM, DevOps, etc.)     (2E13)

[12:30] Mike Bennett: IOF = Industrial Ontology Foundry     (2E14)

[12:30] Ravi Sharma: what is the graph display tool and is it using Sparql?     (2E15)

[12:31] Mike Bennett: If we do want to get manufacturers to use a 'vocabulary' we should extend the idea by having a standard ontology they can use. Concepts > words     (2E16)

[12:32] John Sowa: I just uploaded a file about using CLIP and DOL for knowledge graphs     (2E17)

[12:34] John Sowa: The title of the slides: "Relating knowledge graphs to logic and language" See http://jfsowa.com/temp/kgCLIP.pdf     (2E19)

[12:34] Mark Underwood @knowlengr: Not using Google Schema.org as an SEO specialist borders on malpractice. Having built 40+ sites using that, I can say that is easy to implement ... but (read on)     (2E20)

[12:35] Mark Underwood @knowlengr: data-driven sites that pull product info from databases are not "spiderable," which also needs a bit of coaching for SEO     (2E21)

[12:35] John Sowa: Last week I mentioned the DOL standard and the CLIP dialect for Common Logic. This has more references.     (2E22)

[12:36] Mark Underwood @knowlengr: Images are not the problem; Google insists on these being tagged already - another indicator of lame SEO practice     (2E23)

[12:36] Ravi Sharma: what is the implemented vision state when we reach version one? benefits?     (2E24)

[12:37] John Sowa: Mark, all those issues are addressed by kgCLIP.pdf (in the full version, which I am still working on).     (2E25)

[12:38] Mike Bennett: So this set of categories they have done - is this published somewhere as an ontology in some form?     (2E26)

[12:39] Bobbin Teegarden: DOL standard from OMG: https://www.omg.org/spec/DOL/About-DOL/     (2E27)

[12:39] Mark Underwood @knowlengr: @John - that would be good. To leverage existing web building practice by small businesses, a WordPress plugin would be an optimal enabler     (2E28)

[12:41] David Eddy: Are there definitions for the Keywords?     (2E29)

[12:43] John Sowa: Mark, by "tools", I mean the things that R & D people use to develop the tools that small business could use.     (2E30)

[12:43] Mike Bennett: There are interesting business opportunities that can be enabled by this, e.g. if you design a product from available (even unrelated) parts - effectively defines disintermediated product 'manufacturing'. Could be big.     (2E31)

[12:44] Mike Bennett: Verification can be locked in via Blockchain / Distributed ledger     (2E32)

[12:45] janet singer: @Mike good question re multiplicity of standards and need to classify by concepts of capability, function, etc. Are you saying IOF takes that approach?     (2E33)

[12:46] Mike Bennett: @Janet I believe IOF has made a good start on these concepts. I'm also looking at some of these more deeply for e.g. IoT applications     (2E34)

[12:46] janet singer: IOF work on that related to BFO or not?     (2E35)

[12:47] Mike Bennett: Yes IOF uses BFO as its Top Level Ontology.     (2E36)

[12:48] Ravi Sharma: Binil - thanks for answers     (2E37)

[12:50] ToddSchneider: The ISO web site (https://www.iso.org/obp/ui/#search) shows 124 results when searching for the term 'screw'.     (2E38)

[12:51] Mike Bennett: Screw that.     (2E39)

[12:52] Mike Bennett: That is is why we have ontologies. Unique identifiers should identify a concept not a word. Then terminology is a layer on top of that.     (2E40)

[12:53] Leia Dickerson: Are there slides posted for this talk? I logged in for the tail end.     (2E41)

[12:54] Gary: Lots of "things" with an explosion of parts and sub-types as well as definitions. As with many efforts people start with what they consider an authoritative definition.     (2E42)

[12:55] John Sowa: System programmers are the only humans who should ever look at URIs.     (2E43)

[12:55] David Eddy: @JFS....     (2E44)

[12:55] David Eddy: @JFS... STRONGEST AGREEMENT     (2E45)

[12:56] Mike Bennett: Yes and as with many things they then learn the hard way that you can't do it all from the words and their written definitions. Need classification / subsumption, properties, logical restrictions etc. for the concepts.     (2E46)

[12:57] Bobbin Teegarden: Sigh, so system programmers would be the only ones to differentiate the nuances of what is appearing from e.g. an internet search...?     (2E47)

[12:58] KenBaclawski: @Leia Dickerson: The slides have not yet been posted. I hope to post them soon.     (2E48)

[12:58] Mike Bennett: @Bobbin or it is the challenge for system programmers to relay that knowledge in Human     (2E49)

[12:58] Leia Dickerson: @Ken. Thank you!     (2E50)

[12:59] Bobbin Teegarden: URIs could be (are?) Human...     (2E51)

[12:59] John Sowa: Bobbin. The tools I'm talking about would do the resolutions by talking with anybody or everybody who needs information about anything.     (2E52)

[12:59] Mike Bennett: @Bobbin Have you looked at a Facebook URI for a link to an article?     (2E53)

[13:00] Mike Bennett: We geeks foisted URIs on Humans, I was surprised we ever thought this was a good idea and more surprised we got away with it. Colon slash slash indeed.     (2E54)

[13:01] Bobbin Teegarden: @John similarity of things/answers is a tricky bit, how does Human assess the information they're getting (source, trustability, ...)?     (2E55)

[13:01] Mark Underwood @knowlengr: I liked seeing Neo4j ... was that the first reference in this year's session ?     (2E56)

[13:01] Mark Underwood @knowlengr: Have to jump meetings stay safe all     (2E57)

[13:03] Mike Bennett: Hurrah for segregation of concerns!     (2E58)

[13:03] janet singer: I also need to go thanks Binil     (2E59)

[13:03] Bobbin Teegarden: @JohnS is there a publicly available tool that lets us use CLIP? or will there be?     (2E60)

[13:04] David Eddy: how many people know the OAD is on their iPhone?     (2E61)

[13:04] David Eddy: ... much less care?     (2E62)

[13:04] Ravi Sharma: John - great comment on level of depth URI etc vs system programmers and also on CLIP.     (2E63)

[13:04] David Eddy: actually, BOTH OAD & OED     (2E64)

[13:06] Ravi Sharma: what is OAD?     (2E65)

[13:07] Mike Bennett: TO John's point: Terminology is contextual but an ontology of concepts need not be, within a given scope. But this is getting off topic.     (2E66)

[13:07] Bobbin Teegarden: @JohnS are you saying that CYC with microtheories can create holonic structured ontologies?     (2E67)

[13:07] David Eddy: Oxford American Dictionary, Oxford English Dictionary     (2E68)

[13:07] Ravi Sharma: thanks ken     (2E69)

[13:09] Mike Bennett: I would love to collaborate on this work     (2E70)

[13:09] John Sowa: Bobbin, it's better to say that Cyc (or any truly general-purpose ontology) can support an open-ended variety of contexts.     (2E71)

[13:09] Bobbin Teegarden: @JohnS thank you! yes.     (2E72)

[13:09] John Sowa: I'll send a note about these issues to Ontolog Forum     (2E73)

[13:10] Bobbin Teegarden: @John Please do send.     (2E74)

Resources     (2F)

Previous Meetings     (2G)


Next Meetings     (2H)